Two extant models of thermal folding/unfolding equilibria in two-chain, @-helical coiled coils are tested by comparison with experimental results on excised, isolated subsequences of rabbit @a-tropomyosin ( T m ) . These substances are designated ;Tm, where i a n d j are, respectively, the residue numbers (in the 284-residue parent chain) of the Nand C-terminal residues of the subsequence. One model postulates that a coiled coil consists of segments, each denaturing in an all-or-none manner, like small globular proteins. Thus this model yields a small number of populated molecular species. In an extant calorimetry study of 1lTm127 and of 190Tmzs4, each required only two all-or-none-segments, and their enthalpies and transition temperatures were assigned. These assignments are shown here to yield the concentration of all molecular species, and therefore the helix content, as a function of temperature. Such calculations for lwTm2s4 are in tolerable agreement with CD experiments, but those for 1lTm127 are in gross disagreement. Thus, either the model itself or the calorimetric assignment is faulty. In the second model, all conformational states are counted and weighted, as in the Zimm-Bragg theory for single-chain polypeptides. This theory has been extended (by Skolnick) to two-chain coiled coils and is here used to fit CD data for 11Tmt21, 142Tm281, and 190Tmz&1. The fit is tolerable for llTm127, good for 142Tm281, and quantitative for 190Tm284. Thus this comparison does not falsify this second model. The helix-helix interaction free energy, obtainable from the fit, shows nonadditivity when isolated subsequences are compared with the parent. This suggests that removal of a region from a long coiled coil allows energetically substantial adjustments in side-chain packing in the helix-helix interface. Thus, the helix-helix interaction in long coiled coils is characteristic of a global free energy minimum and not just of the regional constellation of side chains.